文献类型：期刊文献

英文题名：Accelerated nonrigid image registration using improved Levenberg-Marquardt method

作者：Dong, Jiyang[1];Lu, Ke[1];Xue, Jian[1];Dai, Shuangfeng[2];Zhai, Rui[2];Pan, Weiguo[3]

第一作者：Dong, Jiyang

通讯作者：Xue, J[1]

机构：[1]Univ Chinese Acad Sci, 19A Yuquan Rd, Beijing 100049, Peoples R China;[2]Chinese Acad Sci, Inst Remote Sensing & Digital Earth, 20 Datun Rd, Beijing 100101, Peoples R China;[3]Beijing Union Univ, Beijing Key Lab Informat Serv Engn, 97 Beisihuan East Rd, Beijing 100101, Peoples R China

第一机构：Univ Chinese Acad Sci, 19A Yuquan Rd, Beijing 100049, Peoples R China

通讯机构：[1]corresponding author), Univ Chinese Acad Sci, 19A Yuquan Rd, Beijing 100049, Peoples R China.

年份：2018

卷号：423

起止页码：66-79

外文期刊名：INFORMATION SCIENCES

收录：;EI(收录号：20173904202373);Scopus(收录号：2-s2.0-85029708355);WOS:【SCI-EXPANDED(收录号:WOS:000413884200004)】；

基金：The research of this paper is supported by National Natural Science Foundation of China (NSFC, Grant No. U1301251, 61671426, 61471150, 61572077), the Instrument Developing Project of the Chinese Academy of Sciences and Beijing National Science Foundation (Grant No. 4141003).

语种：英文

外文关键词：Medical image registration; Free-form deformation; B-splines; Levenberg-Marquardt optimization

摘要：B-splines have been successfully applied to nonrigid image registration and are popular in various applications. They offer a reduced computational overhead because changes in the control points only affect the transformation within a local neighborhood. Optimization is a key stage in image registration. Most optimization methods only use the gradient direction to determine the update step that may be not optimal. A suboptimal update step may result in a large number of iterations, thus significantly increases the computational time or decreases the accuracy of the registration results. Levenberg-Marquardt (L-M) optimization is a superior algorithm that provides more precise steps during the iteration process. However, because of the large number of parameters in nonrigid image registration, the L-M method suffers from high computational complexity. In this paper, a dedicated optimization method is proposed for nonrigid CT image registration based on L-M optimization. A regular L-M step along with an additional L-M step is computed as the optimal vector, which reduces the computation time because the Jacobian matrix is reused for two calculations in every iteration. Besides, the parameters change automatically according to the calculated results in each step to make the method more efficient. In addition, a linear search for the trial step is introduced to enhance performance. Experimental results indicate that the proposed method is effective and efficient. (C) 2017 Elsevier Inc. All rights reserved.